I am running Rosetta and SETI on my home computer 50/50 share 22/7. Recently SETI started shutting down u/l-d/l during the middle of the week to work on the science behind the Work Units returned.

To enable my computer to span the down time, I increased the Work buffer, although for Rosetta I didn't need it. Also, to avoid continually trying and failing to u/l completed Work Units, I limited network activity (Tu We Th) to 1:00-2:00 AM. The rest of the week is OK anytime.

Since Rosetta is up all week, this early morning window is when I u/l, report, and get new WU from Rosetta and when I am unsuccessfuly doing the same with SETI.

But the Rosetta Work Units have a dreadfully short Report deadline and, possibly because of that, BOINC goes ctazy, scheduling the furthest out first and alternately running all of them, I guess in an attempt to make as many deadlines as possible.

I know you don't work on the BOINC scheduling algorithm, but is there any way I can get Work Units with Report deadline which are 4 or 5 days out, so that BOINC doesn't panic when the deadlines approach? Typical Work Units take 15 to 30 hours to complete, which you can probably see in my profile.

I guess I just feel I'm on too short a leash.

---

I am running Rosetta and SETI on my home computer 50/50 share 22/7. Recently SETI started shutting down u/l-d/l during the middle of the week to work on the science behind the Work Units returned.

To enable my computer to span the down time, I increased the Work buffer, although for Rosetta I didn't need it. Also, to avoid continually trying and failing to u/l completed Work Units, I limited network activity (Tu We Th) to 1:00-2:00 AM. The rest of the week is OK anytime.

Since Rosetta is up all week, this early morning window is when I u/l, report, and get new WU from Rosetta and when I am unsuccessfuly doing the same with SETI.

But the Rosetta Work Units have a dreadfully short Report deadline and, possibly because of that, BOINC goes ctazy, scheduling the furthest out first and alternately running all of them, I guess in an attempt to make as many deadlines as possible.

I know you don't work on the BOINC scheduling algorithm, but is there any way I can get Work Units with Report deadline which are 4 or 5 days out, so that BOINC doesn't panic when the deadlines approach? Typical Work Units take 15 to 30 hours to complete, which you can probably see in my profile.

I guess I just feel I'm on too short a leash.

You could just up your cache instead and then you would have enough Seti work and you wouldn't need to do Rosetta. Rosetta needs the shorter deadlines so it can return work that is relevant to the next phase of testing, work units that take months, not what you want I know, mean that the next phase of testing can't start until they are all back, the ones that didn't crunch properly recrunched, everything testing, the results analyzed, etc, etc, etc. Shorter time frames mean all this can happen quicker and Rosetta itself can move forwards quicker too. If you would like Projects with longer deadlines you can always check here http://distributedcomputing.info/projects.html The projects that crunch under Boinc are noted although the workunit deadlines are not listed. I am not sure I have ever seen that compiled, but it could be.

I am running Rosetta and SETI on my home computer 50/50 share 22/7. Recently SETI started shutting down u/l-d/l during the middle of the week to work on the science behind the Work Units returned.

To enable my computer to span the down time, I increased the Work buffer, although for Rosetta I didn't need it. Also, to avoid continually trying and failing to u/l completed Work Units, I limited network activity (Tu We Th) to 1:00-2:00 AM. The rest of the week is OK anytime.

Since Rosetta is up all week, this early morning window is when I u/l, report, and get new WU from Rosetta and when I am unsuccessfuly doing the same with SETI.

But the Rosetta Work Units have a dreadfully short Report deadline and, possibly because of that, BOINC goes ctazy, scheduling the furthest out first and alternately running all of them, I guess in an attempt to make as many deadlines as possible.

I know you don't work on the BOINC scheduling algorithm, but is there any way I can get Work Units with Report deadline which are 4 or 5 days out, so that BOINC doesn't panic when the deadlines approach? Typical Work Units take 15 to 30 hours to complete, which you can probably see in my profile.

I guess I just feel I'm on too short a leash.

Hi Ross

I haven't received any Rosetta jobs with a deadline of less than two weeks I don't think... I'd suggest you reduce your cache down to 1 day or so, and set your 'connect every' setting to 0.5 days or so which will mean BOINC has very little work queued up and will pick up jobs as it finishes the current ones, leaving you longer to crunch and return them.

If SETI doesn't have work available when BOINC tries to pick it up then it'll do more Rosetta units and then make up the difference when SETI is available.

HTH
Danny

---

Ross, you must have your Rosetta preferences set to indicate a preferred work unit runtime of 24hrs. But you can set this lower if it helps to schedule things more to your liking. Deadlines are 10 days from the point it is downloaded.

Just beware that it will take BOINC several days to get the idea that your work units are running "faster" if you reduce your runtime preference. So it will have some quirks in the first few days trying to schedule the proper amount of work to request.

Eventually it should settle in to where the estimated runtime shown for an unstarted task is roughly equal to your configured runtime preference. If this is not the case now with your 24hr preference, then this may be part of the cause of the scheduling problems you are observing.

---

Rosetta Moderator: Mod.Sense

Thanks. My Rosetta@home preferences show my Target CPU run time as 1 day, so I should be good to go now.

---